Improve connection acquisition timeout error #433
Conversation
Previously driver would propagate `TimeoutException` directly from netty's `FixedChannelPool`. This commit changes it to `ClientException`. Also added couple integration tests.
| @Override | ||
| protected void before() throws Throwable | ||
| { | ||
| assumeTrue( "BoltKit cluster support unavailable", boltKitAvailable() ); | ||
|
|
||
| stopSingleInstanceDatabase(); |
There was a problem hiding this comment.
Do we still have problems to have both single instance and shared cluster running at the same time? This change might make test slower by stopping and restarting server additionally in our tests, right?
There was a problem hiding this comment.
Running both cluster and single instance db leaves only a small amount of free memory on the machine, this makes PR build occasionally hit OOM. I also thought it would make builds significantly slower but it did not, last 3 builds for this change have very similar time as other builds.
| @AfterClass | ||
| public static void stopSharedCluster() | ||
| { | ||
| ClusterRule.stopSharedCluster(); |
There was a problem hiding this comment.
Why we want to immediately shop the cluster after this test class? Should we actually want to keep the cluster running until the whole tests finish?
If this is the only place that requires a cluster, then maybe we should change the ClusterRule to be a class role?
There was a problem hiding this comment.
Cluster takes a lot of memory and might influence other tests. There exists one more test that uses this rule. I can make it a @ClassRule and add cleaning of db in some @After method instead.
Previously driver would propagate
TimeoutExceptiondirectly from netty'sFixedChannelPool. This PR changes it toClientException. Also added couple integration tests.